♟️🎲 Board-Games-Recommender-System 🀄🧩

This repository contains a comprehensive project for a basic board game recommendation system using machine learning clustering techniques. The system leverages the BoardGameGeek database to recommend similar games based on user selection.

---

📑 Table of Contents

1. Overview
2. Project Files
3. Requirements
4. Setup Instructions
5. Usage
6. Conclusions

---

📜 Overview

This project implements a clustering-based recommender system for board games. Users can select a board game and receive recommendations based on similar game attributes such as:

• ⭐ Ratings
• 🎛 Game complexity
• ⏳ Playtime
• 🗂 Game categories and mechanics

The application is built using Python and Streamlit for an interactive user experience.

---

📂 Project Files

This repository contains the following files:

• app.py: The main Streamlit application script to run the recommender system.
• requirements.txt: A list of Python dependencies required for the project.
• ML and Board Games.pdf: A presentation detailing the project objectives, implementation, and conclusions.
• MLClusteringBG.ipynb: The Jupyter Notebook containing detailed steps for data cleaning, exploratory data analysis, data preprocessing, data modeling and clustering analysis for the recommender system.
• boardgamesdf.csv: A preprocessed dataset containing information about board games, including their names, ratings, weights, min/max number of players, and other features. This dataset is generated as the output of the MLClusteringBG.ipynb notebook and serves as the primary data source for the recommender system.

---

⚙ Requirements

To run this project, ensure you have the following installed:

• 🐍 Python 3.8 or above
• 📦 pip (Python package installer)

You can find the specific dependencies in the requirements.txt file.

---

🛠️ Setup Instructions

1. © Clone the Repository:

   git clone <repository-url>
   cd <repository-directory>
   

2. 🌐 Create and Activate a Virtual Environment: It is recommended to use a virtual environment to manage dependencies and avoid conflicts with other Python projects on your system.

   • Create a virtual environment:

     python -m venv env

   • Activate the virtual environment:
     • On Windows:

       .\env\Scripts\activate

     • On macOS/Linux:

       source env/bin/activate

3. 📦 Install Dependencies: Use pip to install the required Python packages:

   pip install -r requirements.txt
   

4. 🗂️ Prepare the Dataset:

   • Ensure the dataset (boardgamesdf.csv) is located in the same directory as app.py.
   • If the dataset is in a different location, update the path in app.py to reflect its location:

     df = pd.read_csv("<path-to-your-dataset>/boardgamesdf.csv")

5. ⚡ Run the Application: Launch the Streamlit app using the following command:

   streamlit run app.py
   

6. 🛑 Deactivate the Virtual Environment: If you no longer need the virtual environment, deactivate it by running:

   deactivate
   

---

🚀 Usage

1. Launch the App: Follow the setup instructions to start the Streamlit application.

2. Select a Board Game:

   • Use the dropdown menu in the app to choose a board game.

3. View Recommendations:

   • The application displays the top 10 rated games similar to your selection.
   • Recommendations are based on clustering analysis of game features.

---

🎯 Conclusions

1. Feature Selection:

   • 🧮 Numerical Features: The final dataset for this project included 6 numerical features: game weight (representing the complexity of the game, on a scale from 1 to 5), min players, max players, playtime, age recommendation, and number of expansions. These were log-transformed, robust-scaled, and min-max scaled to ensure consistency and compatibility with the clustering algorithm.
   • 📂 Categorical Features: The dataset also included 360 categorical features derived from game categories, game subcategories, game mechanics, and game themes. These features were already encoded prior to preprocessing.
   • A mixed features matrix was constructed, combining these scaled numerical features and encoded categorical data, ensuring all data was normalized between 0 and 1 for optimal clustering performance.

2. Clustering Effectiveness:

   • The use of K-means clustering with PCA-reduced data resulted in moderate segmentation of board games into clusters based on features like game weight, playtime, game cathegories and mechanics.
   • Optimal clustering parameters were determined through the silhouette score and elbow method.
   • However, the clustering results revealed large clusters, with up to 8461 games in the first cluster and 2529 in the smallest, making recommendations less precise without additional filtering.

3. Recommendation Accuracy:

   • The system recommends similar board games by leveraging cluster membership.
   • To simplify the recommender system, only the top 10 games within a cluster, sorted by the highest average rating, are displayed.
   • The current recommender system could be further improved by integrating similarity measures such as cosine similarity within clusters to refine results.

4. Scalability:

   • Designed to handle large datasets, ensuring broad applicability for various user preferences.

5. Future Directions:

   • Explore cumulative explained variance to determine the optimal number of PCA components.
   • Experiment with alternative clustering models such as Agglomerative Clustering, DBSCAN, or HDBSCAN to improve cluster precision.
   • Refine the recommender system by incorporating additional steps to measure similarity within clusters and improve overall accuracy.

6. User Experience:

   • Integrated Streamlit for an intuitive and interactive user interface, making game selection and recommendations seamless.